Comment by frenzcan

I agree Asimov's laws are intentionally flawed/ambiguous (which makes the stories so good) but a slight difference to LLMs is the laws aren't just software, the positronic brain is physically structured in such a way (I'm hazy on the details) that violating the laws causes the robot to shutdown or experience paralysing anxiety. So if an LLM's safety rules fail or are subverted it can still generate dangerous output, while an Asimov robot will stop working (or go insane...)